PreFIQs: Face Image Quality Is What Survives Pruning

2026-05-13

AI Summary

This work proposes a training-free and unsupervised method for face image quality assessment. Built upon the Pruned Identification Embedding (PIE) hypothesis, the approach treats the geometric displacement of L2-normalized embeddings—induced by structured pruning of a pretrained face recognition model—as a proxy for image quality. Specifically, it quantifies this displacement via the Euclidean distance between embeddings extracted by the original model and its pruned counterpart, and provides first-order theoretical justification for this design. Relying solely on a pretrained model and Jacobian-vector product approximations, the method achieves state-of-the-art or competitive performance across eight benchmark datasets and four mainstream architectures, demonstrating its effectiveness, generality, and novelty.

Abstract

Face Image Quality Assessment (FIQA) evaluates the utility of a face image for automated face recognition (FR) systems. In this work, we propose PreFIQs, an unsupervised and training-free FIQA framework grounded in the Pruning Identified Exemplar (PIE) hypothesis. We hypothesize that low-utility face images rely disproportionately on fragile network parameters, resulting in larger geometric displacement of their embeddings under model sparsification. Accordingly, PreFIQs quantifies image utility as the Euclidean distance between L2-normalized embeddings extracted from a pre-trained FR model and its pruned counterpart. We provide a first-order theoretical justification via a Jacobian-vector product analysis, demonstrating that this empirical drift serves as a computationally efficient approximation of the exact geometric sensitivity of the latent embedding manifold. Extensive experiments across eight benchmarks and four FR models demonstrate that PreFIQs achieves competitive or superior performance compared to state-of-the-art FIQA methods, including establishing new state-of-the-art results on several benchmarks, without any training or supervision. These results validate parameter sparsification as a principled and practically efficient signal for face image utility, and demonstrate that quality is, in essence, what survives pruning.

Problem

• Face Image Quality Assessment
• Unsupervised FIQA
• Model Pruning
• Embedding Sensitivity
• Face Recognition

Innovation

• Face Image Quality Assessment
• Model Pruning
• Unsupervised FIQA
• Embedding Sensitivity
• Training-Free